Cardiovascular Disease and
Hormone Therapy
Week 2
1
Objectives
1. Learn about cardiovascular disease including sexlinked biology and gender aspects 2. Introduce study designs 3. Learn about the hormone therapy controversy including the centrality of study designs
2
Cardiovascular Disease
•
Class of diseases involving heart & blood vessels
•
Many related to atherosclerosis •
•
Plaque builds up in artery walls
Includes •
Myocardial infarction (heart attack)
•
Ischemic stroke
•
Congestive heart failure
•
Arrhythmias (slow, fast, irregular) 3
4
Cardiovascular Disease in
Women
•
Underlying physiology may be different
•
First myocardial infarction 10 years later •
More likely to die
•
May experience different symptoms
•
Some risk factors more common, powerful
•
Under-diagnosed and under-treated 5
Artery Blockage
Men
Women
6
Disease
Premorbid
Fetal development
Childhood
Birth
Prevention
Onset
Adolescence
Puberty
Adulthood (pregnancy)
Young adult
Primary
Older adult
Menopause
Secondary & Tertiary
Chronic Disease Prevention
Across the Lifespan
7
Fetal development
Childhood
Birth
Adolescence
Puberty
Adulthood (pregnancy)
Young adult
Preeclampsia Gestational diabetes Preterm delivery Low birthweight
Older adult
Menopause
2x risk cardiovascular death
Reproductive Health & Chronic Disease Linkage
8
Vascular dysfunction
Population with complicated pregnancy
Threshold for clinical disease
Healthy Population
Pregnancies
Middle age
Pregnancy as Stress Test for Cardiovascular Disease 9
Satter BMJ 2002
Study Designs Experimental •
Randomized controlled trial (RCT)
Non-experimental or observational •
Case series
•
Ecological/Correlational
•
Cross-sectional
•
Cohort
•
Case-control
•
And many more... 10
Randomized Controlled Trial
11
Randomized Controlled Trial
•
Structure •
Defined by investigator assignment
•
Prospective
•
Measures of association include
•
•
Will cover more next class
•
Risk Ratio, Risk Difference, Odds Ratio
Classic example •
Women’s Health Study •
Tested the effects of lower-dose aspirin and vitamin E in preventing CVD and cancer among 39,876 U.S. female health professionals, over age 45 at baseline
•
Funded by the NIH; based at BWH; industry provided drugs Image courtesy of Keith Ivey on flickr. License CC BY-NC-SA. 12
Randomized Controlled Trial
Strengths include •
•
•
Weaknesses include
Minimizes confounding by known and unknown factors Greater degree of control over exposure Information can be collected on multiple outcomes with little cost increase
13
•
Ethical issues
•
Time consuming
•
Costs and feasibility
•
Must select appropriate exposure, dosing, and duration
•
Compliance, loss-to-follow-up, misclassification
•
Need equipoise
Ecologic/Correlational
14
Ecologic/Correlational •
Structure •
•
Estimate measures include •
•
Information on exposure and/or disease is available on a group level not an individual level
Risk Ratio, Risk Difference, Odds Ratio
Classic example •
Cell phones and brain cancer •
Compare national prevalence of each 15
Ecologic/Correlational
Strengths include
•
Quick
•
Inexpensive
•
Weaknesses include •
Often have a poor measure of exposure
•
No information on if the “exposed” are getting the disease
•
Aggregate association may not reflect individual level association
•
No data to control individual level confounding
Large sample
16
Cross-Sectional
17
Cross-Sectional •
Structure •
•
Estimate measures include •
•
Data on individual level, exposure
and outcome reflect same time
period
Risk Ratio, Risk Difference, Odds
Ratio
Classic example •
National Health and Nutrition
Examination Survey (NHANES)
•
Started 1960s, series of surveys
•
Based at CDC 18
Cross-Sectional
Strengths include •
Quick
•
Inexpensive
Weaknesses include •
19
Can’t access temporality
Cohort
20
Cohort
•
Structure •
Select subjects on the basis of exposure status
•
General or special exposure
•
Prospective or retrospective
•
Estimate measures include •
•
Risk Ratio, Risk Difference, Odds Ratio
Classic example •
Nurses’ Health Study 21
Cohort
Strengths include •
•
Weaknesses include
Observing people as naturally conduct lives
•
Time consuming
•
Expensive
•
Difficult for rare diseases
Recall bias eliminated
•
Good for rare exposures
•
Establish temporality
•
Can estimate risk (unlike casecontrol)
22
Case-control
23
Case-control
•
Structure •
Select subjects on the basis of disease status
•
Retrospective
•
Effect measures •
•
Odds ratio
Classic example •
Doll and Hill’s smoking and lung cancer study
Image courtesy of Roman Pavlyuk on flickr. License CC BY.
24
Case-control
Strengths include
Weaknesses include
•
Fast
•
Difficult for rare exposures
•
Good for rare outcomes
•
•
Short follow-up is ideal for acute epidemic outbreaks of short duration
Can only study one exposure/outcome relationship
•
Limited exposure information
•
Selecting appropriate
controls challenging
25
Study Design Overview
Case series •
Careful, detailed report of a series of patients, highlight factors that could be related to outcome
Randomized controlled trial •
Structure of cohort study, but exposure is allocated by investigator
Correlational (ecologic) study •
Data from entire populations to compare disease frequencies among different groups during the same period of time, or among the same population at different times
Cross-sectional study •
Snapshot in time: information on exposure and outcome of individuals assessed at the same point of time for all subjects
Case-control study •
Observational study with selection into study on basis of outcome status
Cohort study •
Observational study with selection into study on basis of exposure status
26
Study Design Exercise
1. For each description below: Identify the study design used and indicate the main feature that led you to choose that study design. Study design options include: •
case series
•
randomized controlled trial
•
correlational (ecologic) study
•
cross-sectional study
•
case-control study
•
cohort study 27
Study Design Exercise
a. In 1980, an investigator noted that there was substantial variability in per capita fat consumption among 25 European countries. The investigator then also assessed the 1980 coronary heart disease mortality rates in these countries in order to determine whether an association between per capita fat consumption and coronary heart disease mortality in these countries exists. Correlational (ecologic): data are collected on population-
level, not individual-level
28
Study Design Exercise
b. In a study of menstrual abnormalities in females after treatment for childhood cancer, the investigators are enrolling two groups of women who were treated for childhood cancer between 1974 and 1980: (1) women who were treated with chemotherapy and (2) women who were treated with surgery. The frequency of menstrual abnormalities occurring from the
time of treatment through the end of 2004 will be evaluated.
Cohort: comparing a group who was exposed (surgery) to a
group who were not exposed (not surgery, chemotherapy)
29
Study Design Exercise
c. In a study of electric blanket use during pregnancy and its effect on miscarriage, women who are hospitalized for a clinical miscarriage and an age-matched sample of women who are hospitalized for the delivery of a live born infant are being enrolled. All subjects are being interviewed to determine their pattern of electric blanket use during the pregnancy that
just ended. Case-control: comparing a group with the outcome
(miscarriage) to a group without the outcome (live born
infant)
30
Study Design Exercise
d. A physician at MIT Medical is concerned that a high level of self-perceived stress during college is a risk factor for a subsequent clinical diagnosis of depression. She plans on reviewing all of the MIT Medical records in fall 2015. She will identify a group of students who have had a clinical diagnosis of depression, and ask these students
about their previous self-perceived stress levels. Case-series: describing a series of patients with the
outcome, with no comparison group
31
Study Design Exercise
e. A researcher hypothesizes that practicing Tai Chi may lower rates of falls among elderly individuals. She enrolls 1,000 individuals aged 65 years or old and assigns half of them to a Tai Chi program and half of them to usual activities. She then compares the two groups with respect to their rates of falls in the next two years. Intervention: exposure (Tai Chi) was assigned, not self-
selected
32
Study Design Exercise
2. Dietary flavonoids, found in chocolate and plant-based foods, are associated with improved cognitive performance A researcher hypothesized that chocolate consumption may improve not only an individual’s cognitive performance, but also the performance of whole populations •
However, measures of cognitive performance of entire populations are not publicly available for his analysis
He decided to use the number of Nobel laureates per capita in 22 countries as a surrogate marker for cognitive functioning of the population He also obtained information on the per capita yearly chocolate
consumption for these same 22 countries
33
Study Design Exercise
The main figure
from his analysis
The researcher concluded there “is a surprisingly powerful correlation between chocolate intake per capita and the number of Nobel laureates in various countries.”
Messerli, M. D., Franz H. "Chocolate Consumption, Cognitive Function, and Nobel Laureates." . New England Journal of Medicine 367, no. 16 (2012): 1562-4. © Massachusetts Medical Society. All rights reserved. This content is excluded from our Creative Commons license. For more information, see http://ocw.mit.edu/help/faq-fair-use/. 34
Messerli NEJM 2012
Study Design Exercise
a. Discuss three possible explanations for why the authors could have observed an association between chocolate consumption and the number of Nobel laureates from a country. Chocolate consumption influences the number of Nobel laureates. Chocolate consumption has been associated with improved cognitive function and this improved cognitive function could lead to more Nobel laureates. Nobel laureates influence chocolate consumption. People who win Nobel prizes may be more likely to consume chocolate because they are aware of the positive health benefits of chocolate consumption; celebratory events associated with a citizen winning a Nobel prize may increase national chocolate consumption. The number of Nobel laureates and the per capita chocolate consumption are both influenced by a common underlying mechanism. Socioeconomic differences or geographic and climatic factors may explain the association. For example, those countries with higher chocolate consumption may also have higher per capita income which could be associated with strong educational systems. Stronger educational systems should result in more Nobel prize winners. 35
Study Design Exercise
b. Discuss the limitations to the interpretation of the data from this study that are inherent in an ecologic/correlational study. The data in this paper are collected at the national level and we do not have individual level data. We are unable to determine if those citizens who consume the most chocolate are also the citizens who are awarded Nobel prizes. The author only has information on the average amount of chocolate consumed by citizens of each country. We do not know if everyone in that country is consuming the average level of chocolate. The authors are unable to control for confounding by other variables (for example, age or socioeconomic status). Finally, the author does not have information about the timing of chocolate consumption and the awarding of Nobel prizes. We do not know if these levels of chocolate consumption reflect consumption prior to Nobel prizes being awarded. 36
Hormone Therapy
Should women take HT?
Which women?
Which HT?
When? How long?
37
Indications •
Hot flashes
•
Night sweats
•
Vaginal dryness
•
Urethritis
•
Osteoporosis 38
Estrogen Levels
Hormone Therapy
Age 50 39
Feminine Forever •
Defines natural human condition as a disease
•
Cure: “off-label,” unapproved use of a drug that healthy people would take every day for the rest of their lives
•
Proselytizes can accomplish more than symptom relief
•
Receives payments for the book/speaking tours from pharma 40
Endometrial Cancer •
•
Estrogen alone (unopposed) •
5-y use: 4-5 fold increase
•
10-y use: 10-fold increase
Estrogen + progesterone (opposed) •
No association
•
Reason to oppose estrogen
•
Rare: ~55,000 cases diagnosed in U.S. in 2015 41
Cardiovascular Disease
•
•
Meta-analysis of 40 observational studies •
Ever vs. never HT use: RR=0.65 (95%CI
0.59-0.68)
•
Current use: RR=0.50 (95% CI 0.45-0.56)
Common: 1 in 3 women die in U.S. in 2015
42
Grodstein and Stampfer Prog Cardiovasc Dis 1995
Nurses’ Health Study Nurses’ Health Study I (NHSI) •
121,701 female nurses
•
30-55 years of age (1976)
•
Married
Nurses’ Health Study II (NHSII) •
116,609 female nurses
•
25-42 years of age (1989)
Mailed biennial questionnaires Cooperative, >90% follow-up Medical knowledgable -> accurate Homogenous education, career, and race 43
Confounding
Confounder
Hormone Therapy
Cardiovascular Disease
Users vs non-users:
Leaner
Smoke less
More physically active
More educated
More likely to see physician
44
Bernadine
Healy
Head of the National
Institutes of Health &
American Red Cross
Launches $625 million
Women's Health Initiative
Image courtesy of National Institutes of Health Library on flickr. License CC BY-NC-SA.
45
Women’s Health Initiative
Established in 1991, 8-12 year intervention Multi-center randomized controlled trial, UW lead 161,809 women, aged 50-79 Three main areas •
Hormone therapy and
cardiovascular disease
•
Fat intake and breast cancer
•
Calcium/vitamin D and osteoporotic fractures
Image courtesy of the Women's Health Intiative. This image is in public domain. Source: Wikimedia Commons.
Largest randomized trials to date 46
Study Design
Women
Uterus
Estrogen
Without Uterus
Estrogen + Progesteron
Estrogen
47
Placebo
Trial Stopped
•
Study participants informed twice about slight excess risk for CVD among hormone therapy users
•
In 2002, prematurely stopped the estrogen + progesterone component after 5.6 years of followup
•
In 2004, prematurely stopped the estrogen only component after 7 years of follow-up 48
Trial Stopped
Cases
Hazard Ratio (95% CI)
Venous Thromboembolism
218
2.11 (1.58-2.50)
Stroke
212
1.41 (1.07-1.85)
CHD
286
1.29 (1.02-1.63)
Breast Cancer
290
1.26 (1.00-1.59)
49
WHI Investigators JAMA 2002
Prescriptions Decrease
50
WHI Investigators JAMA 2004
How can we explain
the discordant findings from
observational studies and
randomized clinical trials?
51
Confounding •
Users in observational studies may be healthier
•
•
Unmeasured & residual confounding
Users in trial are randomly assigned •
No confounding
52
Trial Non-compliance % non-compliance
50
37.5
25
12.5
0 1
2
3
4
5
6
7
Year of follow-up
53
WHI Investigators JAMA 2002
Observational vs Trial
Results
CHD Stroke Pulm. Emb. Hip Fractures Breast Cancer Colorectal Cancer 0
0.5
1
1.5
2
2.5
3
3.5
4
Observational Trial 54
Michels and Manson Circulation 2003
Different Populations
Trial
Observational •
Elected to use HT
•
Presumably a considerable proportion has menopausal symptoms
•
Started HT when they reached menopause
55
•
Willing to start taking HT at the flip of a coin
•
Had no or only mild menopausal symptoms
•
Started taking HT several
years into menopause
•
70% overweight
•
Possibly unhealthy lifestyle (29% also in diet trial)
CHD in Observational
2.5
Hazard Ratio
2
1.5
1 0.5 0